Javascript must be enabled for the correct page display

Multi robot SLAM pose estimate enhancement

Koeslag, H. (2007) Multi robot SLAM pose estimate enhancement. Master's Thesis / Essay, Artificial Intelligence.

scriptie_1feb08_printversion.pdf - Published Version

Download (1MB) | Preview


Simultaneous localization and mapping, also known as SLAM, is the problem of map making by autonomous robots in an unknown environment given the uncertainty in all sensor readings. This thesis shows how SLAM solutions can be extended by using multiple robots, improving the pose estimates of the individual robots. As already implied by its name, the SLAM problem is twofold. Because the environment is unknown to the robot, localizing itself is a very complex task. On the other hand, without knowing where the robot is located, map making is nearly impossible. The SLAM problem is generally solved by estimating the relative location of the robot since the last sensor sweep based on the odometry, the device(s) that estimate the robot motion. By doing so, the robot has an indication of the location of its current sensor values relative to the previous sensor values. Iterating this process for an extended period of time, accumulating all sensor readings then allows the robot to generate a map of the environment. However, the odometry can merely estimate the robot motion. The same goes for all sensor devices that can be mounted on a robot, they can only make estimates regarding the environment. A way of dealing with these uncertainties is by making nondeterministic pose estimates regarding the robot location, describing the probability of the robot being at a certain location. The pose estimate of an individual robot performing a SLAM task can be described by a probability density function, describing a probability to every location. For a multiple autonomous robot SLAM task, this means that every robot carries a unique probability density function regarding its current pose and a corresponding map of the environment. Whenever robots detect each other and communicate their beliefs about the world, such probability density functions can be combined, narrowing the hypothesis space for each robot. When combining the probability density functions for multiple robots describing their poses, additional information becomes available to the individual robots. This additional information in combination with the detection of the other robot can be used to increase the accuracy of the robot pose estimate. This thesis describes how expanding a SLAM solution with additional robots can increase the quality of the solution. In order to demonstrate the increased accuracy of a robot pose estimate, an existing SLAM framework, DP-SLAM, has been adopted and expanded to facilitate multiple robots.

Item Type: Thesis (Master's Thesis / Essay)
Degree programme: Artificial Intelligence
Thesis type: Master's Thesis / Essay
Language: English
Date Deposited: 15 Feb 2018 07:28
Last Modified: 15 Feb 2018 07:28

Actions (login required)

View Item View Item